Tip:
Highlight text to annotate it
X
Robert Epstein: Good morning. It's a pleasure to be here today.
I actually thought it was a pleasure to be here until Ken spoke and said he wasn't born
in 1976.
[laughter]
Jesus Christ. I mean I was in a rock band. [laughs]
[laughter]
So I feel pretty bummed out frankly.
[laughter]
So it's just a downer from that point onwards. That's all I can say.
[laughter]
I'm a physician epidemiologist and I work in the crap world, as Howard said, of translation.
So, like Helen, I'm definitely not deep in genomics, but I do know a lot about how we
can twist and turn our database to be useful in the genomic space and that's what I was,
I think, asked to come here and speak about today. I'm not even going to trust this thing,
so let me go to page down, I think. So what I'd like to do is just two things in the 10
minutes that I'm here for. Firstly, give you some sense on how you can use payer-type data
to either confirm some hunches that many of you may have by things like phenocopying and
I'll show you one quick example, or investigating potential clinical utility for various tests
and then secondly I would like to pose a point around how the data or the system in which
drugs are paid for in this country is really an interesting and scalable solution to promulgate
the use of these tests where people might believe that the evidence is there.
So, this is a busy slide. I'm sorry. It's the only busy slide that I have, but you know,
for those of you who are curious what is payer data? What are we talking about? I don't actually
use that term. It was given to me for today's presentation. But in the United States, typically
these are the buckets of data that are available within different payer environments and I
work for Medco. We cover 65 million Americans for outpatient drugs, so we have all of their
outpatient drug information. We carry, like Helen was describing, what we call an eligibility
file, which is an anonymized, but linkable individual number for each person, including
their aliases as they switch jobs and things. With that it gives us age, gender, household
relationships, comorbidities in some cases, and it's updated every month so we know if
someone becomes ineligible for some reason and so therefore we wouldn't find any followup
information. We have all their insurance information, all of their claims data, meaning every outpatient
prescription drug that they've received for many, many years.
So we would know things like compliance, switching, nonpersistence, dosing duration. Also, information
about the prescriber, specialty, year since training, which may have some interest to
you if you're trying to figure out who is it that's ordering these genomic tests or
who isn't. Medical claims, inpatient or outpatient with ICD-9 codes, so we know, you know, as
long as it's coded what the reason for hospitalization or some event would be and the presence or
absence of a test, as long as it's coded, but not usually the value. Now, different
than others at Medco, we also have some genomic information and I'll give you a sense for
that in a moment. So, a structured kind of retrospective study in our world, retrospective
meaning not going and approaching patients or doing an IRB-type informed consent.
With our anonymized data you could ask yourself a question for drug X out of 10,000 people
taking drug X, who got a PGX test and who didn't, if you think one of them is ready
for primetime or you think it's being used out there in the real world. Then you could
ask yourself well, on the basis of looking at the physician behavior, did they do something
about it? And those who did or didn't use the test, did they select a different drug
at a different rate? Did they dose differently? Did they perhaps, like in the case of hepatitis
C, pay more attention to duration of therapy? Did the patient's behavior change? We have
a study today underway where we're actually providing the data back to the patient to
see if they're more compliant once they know they are genetically at higher risk for an
adverse problem, actually a clinical problem. Is the use of a test associated possibly with
an increase or decrease of emergency room visits and why? You know, there might be some
hypothesis you'd have about reducing some side effect or bleeding or whatever it might
be, hospitalizations, maybe the use of other tests, or changes in therapy and costs. So
those are the kinds of things you can set up.
I'll give you one example, which you may be aware of. About three or four years ago we
had been reading the literature about clopidogrel and at that time there had been some pharmacokinetic
studies that had suggested that 30 percent of people are either intermediate or poor
metabolizers with 2C19. They were not able to active clopidogrel. That was all very interesting,
but there were really no large scale studies looking at what does all that mean. We have
about a million clopidogrel users at Medco, so for us, if 300,000 of them are intermediate
or poor metabolizers and not getting the benefit of the drug, we need to know about that, because
there might be something we would recommend to them, maybe one of the newer therapies
that have hit the market, et cetera. So, we went into our claims data with some researchers
from Indiana University in this case, looked at new starts of people on clopidogrel who
had had a recent ACS event, compared people who were on clopidogrel by itself for a year
versus those are taking a concomitant very potent 2C19 inhibitor for that same time window,
and just looked to see the MACE events, the cardiovascular end points, and found about
a 50 percent increase relative risk amongst those who were taking a potent 2C19 inhibitor.
So, you know, obviously not a randomized trial, but was 17,000 patients and we did adjust
for, you know, everything on earth and subsequent to our presentation, American Heart Association
and publication, there's been about a dozen studies that are similar with different kinds
of payer databases like these that were in JAMA from the VA looking at the same question
and coming out with somewhat similar odds ratios.
Another question that I heard yesterday, which was around laboratory reporting back to physicians
and what do they do with the reports? We did do a study with the Mayo Clinic Reference
Laboratory and one of the questions we asked ourselves was well, if you know, 500 or 1000
patients -- physicians get back a report on warfarin genotyping, which says something
like you know, patient's got a very high sensitivity, would they in fact reduce the dose or are
they just going to stick it in the chart or look at it and go I don't know what that means
or not do anything? So here we actually linked up data on close to, I think 600 or 700 patients
who were genotyped with warfarin with their subsequent prescription claims files and what
we found was really nicely kind of a dose-dependent change in the weekly warfarin, just the way
the genotypes would've suggested. So, they obviously got the report, we know that, within
three weeks of these prescriptions. And if you look at the very bottom, you see the people
who had very high sensitivity to warfarin had about a 17 milligram per week drop in
the use of their warfarin and at the very top the people who needed a higher dose of
warfarin, in fact, did get a higher dose of warfarin.
So, physicians who got the report, which was generated by the Mayo Reference Laboratory,
which provided this extra layer of interpretation, physicians actually did change behavior based
on the report in a way that you would've expected. So I'd say, you know, some of the buckets
of things then that we could look at are does genomic test information result in differences
in compliance like fear factor? You know, can you scare people based on their use -- or
their genomic information to stay on their therapy? Because in our world we find with
most chronic meds people -- about 50 percent of people drop off in the first year and that's
despite trying to explain to them how important it is to stay on their therapy.
One little twist in the theme may be that genomic information would be that extra piece
of information that scares them enough to stay on the therapy. So we're actually looking
at that today. I share with you the example on physician behavior change and also examples
on major clinical events and resource utilization. Lots of limitations with payer data and we
could go on to the conversation for three hours, but you know, here's just some of them
in full disclosure. I think the biggest one, just in terms of -- from an analytic perspective,
in the U.S. anyway, the medical claims part, the hospitalizations and all of that, those
get looked at and kind of fooled around with and readjudicated for upwards of five months
so they're not fresh the way drug data are, which are instantaneous every day.
Switching gears quickly. Yep, two minutes. One of the big problems with promulgating
testing beyond figuring out about the evidence is physician awareness in the field. We did
do a 10,000 physician survey with the American Medical Association a couple of years ago
and these data are in press currently. Good news on the far left is that 98 percent of
American physicians do believe that genes do relate to drugs, so that's a good thing,
because if they didn't believe that we'd really be in trouble. But if you look at the arrow
and the bar next to the arrow only about 10 percent of physicians feel they know enough
to even order a test. Ninety percent don't feel comfortable, don't remember having any
training or education in genetics. So that's a real problem.
And so the one solution I would throw out there that is scalable relates to the only
part of the U.S. health care system that's 100 percent wired today, which is pharmacy.
Since 1990, all 60,000 outpatient pharmacies in America have been electronically wired
with the same data elements being entered no matter where you go. So for the 90 percent
of Americans who carry an insurance card in their wallet, if you go to Hawaii one day
and New York the next we're going to know it and if there's a drug-drug interaction
it pops up right on the screen with decision support rules, you know, right there to the
pharmacist or to the physician if they use an E prescribing device. What we've been doing
at Medco the last few years is using that same system to propagate a rule that says
did you know there is this genomic test for this drug? You might want to consider it.
The payer is paying for it. Or if we have the genomic data, did you know there's a gene-drug
interaction and here it is and warning the physician or the pharmacist about it. So,
this is a scalable way in this country to get on board an existing wired system with
gene-drug rule sets that don't exist today for systems like this that are resident in
every single pharmacy in America. Of course it's all about partnerships and collaborations,
public and private. That was the point for those of you who were not born before 1980
--
[laughter]
This was President Nixon and Elvis Presley. You laugh, but I was at a meeting about a
year ago using the same slide and a young person came up afterward and said, "Who are
those two people?"
[laughter]
So these are some of the folks we've been working with. I just mostly want to acknowledge
the very bottom row, which is the more than 150 payers and more than 50,000 patients who
have participated in one or more studies with us to date. And these are my conclusions.
I am going to stop there and take any questions. Thank you.
[applause]
Fire away.
Male Speaker: So, on your warfarin example, what I'm wondering
is how you get the other data besides the drug information? So, for example, how do
you know that the physicians didn't change the dosing because they were doing clotting
times?
Robert Epstein: Oh. Well, we have the INR tests, whether they
do or don't do them. So we would know if they did or didn't have an INR test between the
date of their genomic test and the date of the drug change. We wouldn't know the INR
time, I mean the actual result, but we would know the presence or absence of having the
test.
Male Speaker: So you know they didn't do --
Robert Epstein: Correct.
Male Speaker: Those tests.
Robert Epstein: That's right.
Male Speaker: Oh. Well, I'm surprised, actually, that's
great.
[laughter]
Robert Epstein: You know, what you find in the real world
of data like this is people don't behave the way you think they do, you know, in academic
settings or other places. You know, the frequency with which INR testing is actually done is
pretty infrequent after the first week or two of warfarin therapy. It's kind of sad,
but a lot of things are like that. When we --
Male Speaker: Shouldn't that genetic tests be done for [unintelligible]
therapy the same time they should be --
Robert Epstein: Well, theoretically it would be, yes. You
know, even to -- we went to a panel of oncologists asking them, you know, do you think for imatinib
users, a reminder system on BCR-ABL testing would be a good idea, because it's in everybody's
guidelines to do that a certain frequency. And they all said no, no. We don't need to
be told, because we know to do it blah, blah, blah. We looked at the data set and, you know,
40 percent of the time it's not being done, nationally. So you know, maybe in your practice
you do it, but it's not happening. Yep.
Female Speaker: You said under most circumstances you're not
--
Robert Epstein: Oh, sorry.
Female Speaker: Sorry.
Robert Epstein: Sorry, sorry, sorry.
Male Speaker: Could you comment -- so you manage this, but
you have people you report to, companies and employers of various sorts. What's their level
of interest in genetics?
Robert Epstein: That's a great question. So, about two years
ago we actually surveyed 700 payers who represent about 60 billion dollars of prescription drug
expenditures in this country and we had them force rank a bunch of different topics and
pharmacogenomics was one on the list. It's the number two topic of interest. Do you believe?
So, their number one is like benefit designs, co-pays, deductibles, that kind of stuff.
But number two was pharmacogenomics. That beat out generics, bio-tech drugs, pipeline,
you know, all sorts of drug-related questions, but pharmacogenomics has the payer's interest
right now. They're looking at it as a solution, not so much as an obstacle, but a solution
to the imprecision that happens today.
Male Speaker: Could you comment on --
Male Speaker: [unintelligible]
Male Speaker: Oh, sorry.
Male Speaker: I have a quick question. I was interested
in the low prevalence or the low comfort level of physicians in ordering tests and you're
talking about delegating to pharmacists the advising on the genetic testing, so for you
specifically and maybe more generally for this group is, what's the role of the medical
geneticists and genetic counselors in all of this? And I think it's a question specifically
here it would be real interesting to hear the answer, but just in general, you know,
how's that community which has traditionally been the experts in counseling going to interact
with all this new data?
Robert Epstein: Let me first say that there does not seem
to be a natural home for who owns the communication and understanding around all this new science
and there needs to be. I mean, there's lots of different players in this space. We hire
in -- we have about 30 genetic counselors on our staff today who talk to physicians
every day, but that's not a scalable solution. There are only 2000 of them in the country,
so you know, who should own this information and be the person communicating it? And I
don't have a great answer for that one. I've been to so many meetings where everybody says
you know, it's over there, but I'm not sure about the answer to that question. But we
do need somebody who really owns and keeps people up to date and has a system that learns
and educates people, because the knowledge is moving so quickly it's very hard for people
who are in practice to keep up.
Male Speaker: And by the way, I recognize Nixon, but I was
wondering who that person on the right was.
Robert Epstein: [laughs]
[laughter]
Male Speaker: So, could you comment on Medco's DNA Direct
company and sort of how you think for something that needs a scale at this level, does it
need to be a private entity who can hire 40 people to work on this or can -- should it
be a government entity? Could you comment along those lines?
Robert Epstein: Sure. We -- DNA Direct was a company we acquired
a couple of years ago. We started off doing lots of studies and research and helping payers
get access to individual tests that were related to drugs, of which there may be like 40 total.
But what the payers told us was the genetic testing in general was so much bigger. You
know, you know this. There's 2000 to 3000 genetic tests and they wanted help managing
all of that. So, we identified a company called DNA Direct who had sort of a system and software
that helps physicians do decision support and figure out, you know, are the indications
right for that given test? An example of where a finding in that product where it's been
used, and I won't say the percent, but a surprising percent of physicians order a BCR-ABL test
when what they really wanted was a BRCA1 test and vice versa and I'm not kidding. So you're
getting a ton of leukemia tests in breast cancer people and vice versa because they
start with B.
[laughter]
Well, it's the way it's working. I'm serious from real experience. So, you know, we do
need some kind of decision support tools and I don't know if it has to be the private sector,
the public sector, but we definitely need to get ready for the future.
Male Speaker: [unintelligible]
Male Speaker: I think we stop there and move on.
Robert Epstein: You can ask me questions during the break.
It's fine, if you want. Thank you very much. Appreciate it.
[applause]